Title
Author(s)
Citation
Issue Date
URL
Rights
Production benefits of childhood overhearing
Knightly, LM; Jun, SA; Oh, JS; Au, TKF
Journal of the Acoustical Society of America, 2003, v. 114 n. 1, p. 465-474
2003
http://hdl.handle.net/10722/42604
Production benefits of childhood overhearing Leah M. Knightly Department of Psychology, University of California at Los Angeles, Los Angeles, California 90095
Sun-Ah Juna) Department of Linguistics, University of California at Los Angeles, Los Angeles, California 90095
Janet S. Oh Department of Psychology, University of California at Los Angeles, Los Angeles, California 90095
Terry Kit-fong Au Department of Psychology, University of Hong Kong, Hong Kong
共Received 15 August 2002; accepted for publication 7 April 2003兲 The current study assessed whether overhearing Spanish during childhood helps later Spanish pronunciation in adulthood. Our preliminary report based on a subset of the data 关Au et al., Psychol. Sci. 13, 238 –243 共2002兲兴 revealed that adults who overheard Spanish during childhood had better Spanish pronunciation, but not better morphosyntax, than adult learners of Spanish who had no childhood experience with Spanish. We now present data from the full sample with additional morphosyntax and pronunciation assessments, as well as measures to help rule out possible confounding prosodic factors such as speech rate, phrasing, and stress placement. Three groups of undergraduates were compared: 15 Spanish–English bilinguals 共native Spanish speakers兲, 15 late learners of Spanish who overheard Spanish during childhood 共childhood overhearers兲, 15 late learners of Spanish who had no regular experience with Spanish until middle or high school 共typical late L2 learners兲. Results confirmed a pronunciation advantage for the childhood overhearers over the typical late L2 learners on all measures: phonetic analyses 共VOT and degree of lenition兲, accent ratings 共phoneme and story production兲, but no benefit in morphosyntax. Importantly, the pronunciation advantage did not seem attributable to prosodic factors. These findings illustrate the specificity of overhearers’ advantage to phonological production. © 2003 Acoustical Society of America. 关DOI: 10.1121/1.1577560兴 PACS numbers: 43.70.Ep, 43.70.Fq, 43.71.Hw 关AL兴
I. INTRODUCTION
Phonology is difficult for late second-language 共L2兲 learners to master 共Oyama, 1976; Tahta et al., 1981; Flege and Fletcher, 1992; Flege et al., 1995, 1999兲. Late learners’ difficulty in producing nativelike accents in their L2 may in part be due to perceptual deficits 共Flege, 1995兲. In fact, perceptual training in identification of L2 sounds seems to help late L2 learners pronounce the sounds of the target language better 共Bradlow et al., 1997, 1999兲. Childhood exposure to the target language seems to benefit adult L2 perception and perhaps production as well. Case studies 共Wode, 1981; Yamada, 1995兲 suggest that children learning English during visits to the U.S. maintained English-like production and perception after a two-year absence. Tees and Werker 共1984兲 found that English-speaking adults who lived in a Hindi-speaking environment for the first year or two of their lives perceived Hindi phonemic contrasts reliably better than those who had no childhood experience with Hindi. However, it remains unclear what type of language experience 共e.g., hearing, speaking兲 contributed to these perceptual and production benefits. Infants learn much about the phonology of their ambient language simply by hearing it. They form language-specific a兲
Author to whom correspondence should be addressed. Electronic mail:
[email protected]
J. Acoust. Soc. Am. 114 (1), July 2003
vowel categories by 6 months of age and display ‘‘languagespecific phonetic’’ perception by 10–12 months 共Werker and Tees, 1984; Kuhl et al., 1992; Werker, 1995兲. Impressively, 8-month-old infants learn word boundaries in a continuous speech stream after 2 min of exposure, using only statistical cues—where the only cue available for word boundaries is that sounds within a word are more likely to co-occur than sounds across word boundaries 共Saffran et al., 1996; Johnson and Jusczyk, 2001兲. Furthermore, such learning seems to be incidental rather than conscious and explicit. Saffran et al. 共1997兲 asked adults and 6- and 7-year-old children to create pictures on a computer 共cover task兲 while an artificial language played in the background. Even though participants were told they would not be tested on the acoustic material presented 共i.e., the artificial language兲, both adults and children nonetheless demonstrated on a later test that they learned words of the artificial language after only 20 min of ‘‘incidental’’ exposure to it. Saffran et al. concluded that such ‘‘incidental’’ learning could be important in natural language acquisition. Yet, to date few language acquisition studies have examined ‘‘incidental’’ learning per se outside the laboratory setting. Our study tries to fill this gap by focusing on ‘‘incidental’’ language learning in a natural language setting 共i.e., overhearing everyday conversations兲. In an interim report of the current study, we explored possible effects of overhear-
0001-4966/2003/114(1)/465/10/$19.00
© 2003 Acoustical Society of America
465
ing Spanish during childhood 共Au et al., 2002兲. In that report, we compared three groups of English-speaking college students on their acquisition of Spanish phonology and morphosyntax. They included 10 Spanish-English bilinguals whose first language was Spanish 共native speakers兲, 11 native English-speakers who regularly overheard Spanish during childhood but first learned Spanish in class around age 14 共childhood overhearers兲, and 12 native English-speakers who had minimal exposure to Spanish prior to learning it in class around age 14 共typical late L2 learners兲. Compared to the typical late L2 learners, childhood overhearers’ pronunciation of Spanish /!, #, %, ", $, ,/ turned out to be more nativelike, as assessed by phonetic measures and accent ratings. Specifically, their VOT 共voice-onsettime兲 for /!, #, %/ was reliably shorter for word initial position 共e.g., គt acos), thus closer to the norm of Spanish VOT, a short-lag VOT approximately 30–50 ms shorter than the long-lag VOT of English /!, #, %/ 共Lisker and Abramson, 1964兲. Overhearers also produced Spanish intervocalic /", $, ,/, which are typically voiced and lenited 共i.e., fricatives or approximants兲 but remain stops and are sometimes devoiced in English, more often as lenited consonants compared to the typical late learners. Importantly, overhearers’ English pronunciation was not compromised; their pronunciation of English /!/ and /"/ was not reliably different from that of the typical late L2 learners. Interestingly, the benefits of childhood overhearing did not extend to the area of morphosyntax. Au et al. found that childhood overhearers performed no better than typical late L2 learners in detecting morphosyntactic errors in a grammaticality judgment task or in their production of correct number and gender agreement in noun phrases. These interim findings suggest that simply hearing the language during childhood does not seem to benefit the mastery of morphosyntax, which like phonology is also challenging to late L2 learners 共Snow and Hoefnagel-Hohle, 1978; Patkowski, 1980; Johnson and Newport, 1989; Johnson et al., 1996; Flege et al., 1999; Birdsong and Molis, 2001兲. Although Au et al. 共2002兲 found no benefits in morphosyntax, a more comprehensive assessment may yet show an overhearing advantage in this area. Additionally, the benefit in phonology reported by Au et al. could be due to confounding factors of prosody such as speech rate, placement of stress, and phrasing. Voiceless stop consonants tend to have longer VOT in slow speech, stressed syllables, and phrase initial position compared to those in fast speech, unstressed syllables, and phrase medial position, respectively 共de Jong, 1995; Fougeron and Keating, 1997兲. Spanish intervocalic /", $, ,/ often become stops after a pause 共Stockwell and Bowen, 1965兲, and slow speech typically contains more pauses than fast speech. In Au et al.’s study, participants were asked to produce word initial and medial /!, #, %, ", $, ,/ in stressed syllables of Spanish words embedded in the carrier phrase Diga គ por favor 共meaning ‘‘Say គ please’’兲. Overhears could have produced more nativelike 共i.e., shorter兲 VOT for word initial /!, #, %/ if the typical late learners happened to produce the carrier sentence more slowly than the overhearers, or if the overhearers misplaced stress on the Spanish words more often than the typical late learners 466
J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
共thereby producing more target phonemes without stress兲. Similarly, typical late learners’ less frequent lenition of Spanish /", $, ,/ could have resulted from less fluent speech and hence more pauses before the target phonemes. To reevaluate Au et al.’s 共2002兲 interim findings, we report here the originally planned study in its entirety with the full sample, along with more in-depth phonetic analyses 共degree of voicing in voiced stops, measurement of prosodic factors, as well as VOT and stop closure of voiceless stops, and lenition of voiced stops兲, additional accent rating data 共narrative production兲, and additional assessment tasks in morphosyntax 共verb morphology and story telling兲. In sum, the study presented here takes a more comprehensive look at whether there is a childhood overhearing advantage specific to phonology and not apparent in morphosyntax. II. METHOD A. Participants
Undergraduates enrolled in second year Spanish language courses at UCLA were recruited to complete a detailed language background questionnaire and follow-up interview about their experience with Spanish from birth to the time of testing. All participants completed a consent form and were paid for their participation. Out of the 238 participant responses received, 15 were identified as childhood overhearers and 100 were identified as typical late L2 learners. From the latter group, 15 were randomly selected with the constraint that they matched the childhood overhearers by gender and Spanish language instructor. Agreement between two trained coders on group assignment 共childhood overhearer versus typical late L2 learner versus unclassifiable兲 was excellent 共91% agreement; Cohen’s Kappa⫽0.90). An additional 15 UCLA undergraduate native Spanish speakers were recruited for the native speaker group.1 Three groups of 15 speakers 共ten women, five men兲 made up the final sample. Table I presents demographics and several language use measures for these three participant groups. Childhood overhearers and typical late L2 learners were all born in the U.S. and first learned Spanish in middle or high school. The overhearers reported overhearing Spanish during childhood from a parent/relative regularly and being spoken to in Spanish and speaking Spanish minimally 共limited to short phrases and words in Spanish兲. Their amount of self-reported experience overhearing Spanish during ages 0– 6 differed reliably from zero 关 t(14)⫽2.33, p⬍0.05兴 whereas their being addressed to in Spanish and speaking Spanish during ages 0– 6 did not 关 t’s(14)⬍1.58, n.s.兴. Typical late L2 learners had minimal, if any, childhood exposure to Spanish. Native speakers were Spanish–English bilinguals who learned English as their L2 before age 10, and most of them were born in the U.S. Both native speakers and childhood overhearers were of Mexican or Central American descent. Overhearers were mostly mixed-Latino, with Spanish-English bilingual parents and relatives, whereas native speakers were mostly full-Latino, with monolingual Spanish-speaking parents and relatives. Typical late L2 learners were of non-Latino descent, with monolingual English-speaking parents and relatives. Knightly et al.: Production benefits of childhood overhearing
TABLE I. Summary of speaker demographics and language use measures.
First language Age 共years兲 Hrs/wk heard Spanish age 0– 6 age 6 –12 Years taken Spanish classes middle/high school college No. visits to a Spanish-speaking country Reported % use of Spanish 共vs English兲 during high school years during college years Reported degree of 共Phinney, 1992兲 identification with ethnic group participation in Latino practices Slang task performance 共% correct兲
Native speakers
Childhood overhearers
Typical late L2 learners
Spanish 22.3a 共0.7兲
English 19.9b 共0.6兲
33.0a 共0.0兲 30.1a 共2.0兲
5.3b 共2.3兲 1.8b 共0.7兲
0 c (0) 0.13b 共0.09兲
2.4a 共0.3兲 0.9a 共0.4兲 6.7a 共1.5兲
3.9b 共0.3兲 1.0a 共0.2兲 1.5b 共0.5兲
4.4b 共0.3兲 0.6a 共0.1兲 0.1b 共0.09兲
41.7a 共3.5兲 30.4a 共3.8兲
9.3b 共2.7兲 7.5b 共1.5兲
9.9b 共2.9兲 8.0b 共1.6兲
3.9a 共0.04兲 3.5a 共0.1兲 80.9a 共2.6兲
3.2b 共0.1兲 2.9b 共0.1兲 19.8b 共2.4兲
3.3b 共0.2兲 2.4c 共0.2兲 2.7c 共0.9兲
English 18.7b 共0.3兲
Note. Table indicates group means with standard errors in parentheses. Means with different superscripts within a row were reliably different from each other according to Tukey’s HSD test, p⬍0.05. Phinney Ethnic Identity Measures 共1992兲 were on a four-point scale with higher numbers indicating more identification and participation.
Childhood overhearers and typical late L2 learners’ selfreports were corroborated by reports from independent informants who knew the participants’ prior experience with Spanish 共e.g., parents兲. Reports on 40% of the research participants confirmed childhood overhearers’ regular passive exposure to Spanish and limited spoken Spanish 共single words, short phrases兲 and typical late L2 learners’ lack of childhood exposure to Spanish. For further corroboration, participants’ knowledge of Mexican/Central American household childhood expressions was tested to assess their childhood exposure to Spanish in the home. Participants read 20 English expressions 共e.g., cry baby, pacifier, dry crust in eyes兲—one at a time—on a computer screen and were asked to translate them into informal Spanish as they heard them at home, in the neighborhood, or in a schoolyard 共slang production兲. They also heard 40 Spanish slang terms 共e.g., chiqueado meaning ‘‘spoiled child;’’ las escondidas meaning ‘‘hide-and-seek’’兲 via a headset and were asked to translate them into English 共slang comprehension兲. Participants’ responses were audio-recorded and later independently transcribed and coded by two research assistants who were native Spanish speakers 共average agreement between coders ⫽96%, disagreements were resolved by a third native Spanish speaker兲. As seen in Table I, the results corroborated participants’ self-reports of childhood experience with Spanish quite well, suggesting that the childhood overhearers knew far less Spanish childhood slang than the native speakers, but nonetheless knew reliably more than the typical late L2 learners.
each word containing the target phoneme in a stressed syllable 共e.g., bគ ase ‘‘base,’’ cabគ eza ‘‘head;’’ see Table II for a complete list兲. Participants were asked to say each of the 36 target words in the sentence frame, ‘‘Diga 共target word兲 por favor,’’ meaning ‘‘Say 共target word兲 please,’’ thus producing 36 target sentences. To see if overhearing Spanish during childhood might compromise speakers’ English pronunciation, participants were asked to produce English voiceless consonant /!/ 共in pគ epper, pគ ocket兲 and voiced consonant /"/ 共in bគ eggar, bគ onnet兲 in the sentence frame, ‘‘Take a 共target word兲 once again.’’ In order to lessen the potential prosodic confounds discussed earlier, participants were asked to stress the target word in each sentence and avoid pausing between words. Participants were given two practice sentences before reading the target sentences. Each sentence was presented three times in random order, and displayed on the computer screen for three seconds. Instructions and stimuli for all tasks were presented on a Macintosh Powerbook G3 or 3400c/200 using PsyScope 共Cohen et al., 1993兲, with auditory stimuli presented via a headset, and participants responding via a button box. Participants were tested individually in a soundproof room, and their utterances were recorded using a Sennheiser HMD 25-1 microphone headset and a Marantz PMD-222 or PMD-430 professional recorder.
B. Phoneme production assessment
Speech recordings were digitized at 12.5 kHz, and phonetic measurements were made using KAY Elemetrics speech analysis programs CSL and MultiSpeech. All measurers/coders 共one primary, two secondary兲 were blind to the speakers’ language backgrounds and analyzed roughly the same percentage of speakers from each of the three groups 共native speakers, childhood overhearers, typical late
1. Stimuli and procedure
As in Au et al. 共2002兲, pronunciation of Spanish phonemes /!, #, %, ", $, ,/ was assessed. Twelve categories were created 共3 places of articulation⫻2 types of voicing⫻2 positions in word兲 with three target words per category and J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
2. Measurement
Knightly et al.: Production benefits of childhood overhearing
467
TABLE II. Spanish word list. Target phoneme
Word initial
English gloss
English gloss
/"/
base beca beso
base scholarship kiss
cabeza jabo´n sabor
head soap taste
/$/
datos deja dı´a
data to leave day
nadar pedido rodar
swim an order roll
/,/
gallo gato goma
rooster cat glue
hogar pago´ pego´
home paid hit
/!/
pase pena peso
to pass/pass embarrassed weight
zapeta vapor tapo´n
diaper vapor stopper
/#/
tacos teja tı´a
tacos shingling/weave aunt
matar metido notar
to kill it’s in to notice
/%/
callo caso coma
a blister case/pay attention command to eat/eat
tocar saco´ peco´
to touch took out sinned
L2 learners兲. To assess agreement, two speakers analyzed by each secondary measurer/coder were randomly selected to be analyzed by the primary measurer/coder.2 a. VOT and stop closure. For /!, #, %/, VOT was measured from the stop release to the onset of the second formant 共F2兲 of the following vowel. Stop closure duration was measured from the offset of F2 of the vowel before the target phoneme to the release of the target phoneme. Mean differences between the primary and secondary measurer共s兲 for individual speakers were minimal for VOT 共ranging from 2.6 to 3.6 ms for Spanish; 2.3 to 2.7 ms for English兲 and stop closure duration 共ranging from 3.9 to 13.9 ms for Spanish; 2.1 to 2.3 ms for English兲. b. Lenition. Spanish voiced consonants /", $, ,/ were categorized as either stops or lenited consonants, distinguished by an abrupt or a gradual change, respectively, in amplitude between the consonant and the following vowel. Percent agreement on these categorical assessments was high, ranging from 91% to 100% 共Cohen’s Kappas between 0.70 and 1.0兲. c. Voicing. To assess the production of voiced consonants more comprehensively, we examined the degree of voicing during the consonant by categorizing all target consonants as having one of three voicing types: voiceless, partial voicing, or full voicing. For stops, voicing type was determined by voicing during the closure; for lenited consonants, by voicing during the duration of the consonant. Tokens were classified as ‘‘voiceless’’ if there was no voicing 共i.e., no voice bar in spectrogram兲, as ‘‘full voicing’’ if voicing was present throughout the entire consonant duration, and as ‘‘partial voicing’’ if they displayed some voicing but not full voicing. Percent agreement on voicing among the coders was high for Spanish, ranging from 82% to 94% 共Cohen’s Kappas between 0.78 and 0.92兲 and for English, 83% to 92% 共Cohen’s Kappas 0.71 to 0.80兲. d. Prosodic factors (Speech rate, phrasing, and stress 468
Word medial
J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
placement). Speech rate was assessed by the duration from the F2 onset of the vowel after the release burst of the first word in the carrier sentence 共e.g., /{/ in ‘‘diga,’’ /|/ in ‘‘take a’’兲 to the F2 offset of the vowel before the target word 共e.g., /~/ in ‘‘diga,’’ /./ in ‘‘take a’’兲. Mean differences between measurers on individual speakers for speech rate were on average 9.4 ms for Spanish and 2.6 ms for English. To measure phrasing differences, a coder listened to each Spanish token and judged whether a phrase boundary had been inserted before the target phoneme 共subjective measure兲. As a quantitative measure a ratio of stop closure duration to speech rate duration was calculated for each token categorized as a stop 共see Sec. II B 2 b兲. Since stop consonants in phrase initial position tend to have longer closure duration, most of the tokens judged to have a prosodic boundary had a ratio around 0.75, and those without around 0.50. Using this criterion, all tokens were reevaluated so that final classification was based on ratio 共objective measure兲; only tokens with a ratio of 0.75 or greater were classified as having a phrase boundary before the target sound. For stress misplacement, a coder listened to each Spanish token and noted whether speakers placed stress on the incorrect syllable 共e.g., correct—PAse, incorrect—paSE兲. C. Accent ratings
As in Au et al. 共2002兲, we asked another group of native Spanish speakers to rate the participants’ pronunciation of the target phonemes /!, #, %, ", $, ,/ in the target sentences 共phoneme accent ratings兲. To assess their accents in more natural speech, we asked yet another group of native Spanish speakers to rate participants’ accents in narrative production 共narrative accent ratings兲.3 a. Phonemic accent ratings. Forty-eight native Spanish speakers4 were recruited to rate participants’ pronunciation of the target phonemes in the 36 target sentences 共e.g., ‘‘Diga Knightly et al.: Production benefits of childhood overhearing
pគ ase por favor.’’兲. The second of the three tokens produced by each speaker for each of the 36 target sentences was selected for rating. To avoid rater fatigue, 36 blocks were created whereby each block consisted of the same target sentence spoken by all speakers. Blocks were arranged into three sets so that each set included 12 blocks of target sentences containing the six target phonemes 共/!, #, %, ", $, ,/兲 in word-initial and medial position. In this way, three tokens of a category 共e.g., tokens pase, pena, peso of word-initial /!/ category兲 were dispersed among the three sets 共see ‘‘stimuli and procedure,’’ Sec. II B 1兲. Each rater was asked to rate each speakers’ pronunciation of the target sound 共e.g., the /!/ in pase兲5 for one of the three sets. Prior to rating, raters listened to each speaker say, ‘‘Diga teja por favor,’’ to familiarize themselves with the range of speakers’ pronunciation abilities. A rating scale was presented on the computer screen during the familiarization and accent rating trials 关 1⫽very strong foreign accent, definitely non-native; 2 ⫽strong foreign accent; 3⫽noticeable foreign accent; 4 ⫽slight foreign accent; 5⫽no foreign accent, definitely native; adopted from Bongaerts et al. 共1997兲兴. Sentences within blocks were randomized and the test was self-paced. Interrater reliability was assessed using all of the ratings to compute, for each rater, an average rating for each speaker. An intraclass correlation on these averages revealed excellent agreement 共average accent scores for individual participants: intraclass R⫽0.98, p⬍0.0001). b. Narrative accent ratings. To assess participants’ accents in more natural speech, we elicited narratives using a 10-page abridged version of a wordless children’s picture book in Mercer Mayer’s ‘‘Frog, Where are you?’’ series—a widely-used task for eliciting narratives from children and adults in different languages 共e.g., Berman et al., 1994兲. Participants were given 2 min to scroll through the storybook pictures on a computer before coming up with a story. They then saw each page for 12 s, allowing time to say one or two sentences in Spanish per page. Audio recordings were independently rated by two native Spanish speakers6 using the same scale adopted for phonemic accent ratings. The intraclass correlation between the two raters was excellent (R ⫽0.91). D. Morphosyntax assessment
In addition to the grammaticality judgment task and noun-phrase production task reported in Au et al., the narrative production task just described and a verb-phrase production task were included to yield a more comprehensive picture of participants’ mastery of Spanish grammar. For all tasks reported in this section, stimuli are available upon request. a. Grammaticality judgment task. Participants listened to 33 grammatical–ungrammatical sentence pairs spoken by a native Spanish speaker. They heard the sentences in random order; they heard each sentence twice and pressed a button to indicate whether it was grammatical or ungrammatical. Both decision and response time were recorded. To minimize fatigue, the 66 sentences were presented in two blocks intermixed with other tasks. Ungrammatical sentences contained an error in one of the following categories: number/gender J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
agreement in noun phrases 共e.g.,*la flores,*el carro blanca兲, number agreement in verbs 共e.g.,*Marta corren,*mi mama´ toman兲, tense-aspect marking in verbs 共e.g.,*Dentro de cua˜ os,*soy un abogado兲, negation 共e.g.,*El conoce a tro an ˜ a a nosotros兲, or person nadie兲, indirect object 共e.g.,*El ensen agreement 共e.g.,*nosotros comienzan,*A que´ hora llegue´ usted兲. b. Noun-phrase production task. Participants were asked to verbally complete five simple four-piece jigsaw puzzles designed to elicit four combinations of number and gender markers 关adapted from Plann 共1979兲兴. Each puzzle appeared on the computer screen for 18 s with four puzzle pieces and a puzzle frame 共showing numbered spaces for the pieces兲. For example, pieces in one puzzle depicted two white pianos 共los pianos blancos: plural masculine兲, two white cows 共las vacas blancas: plural feminine兲, a black piano 共el piano negro: singular masculine兲, and a black cow 共la vaca negra: singular feminine兲. To complete the puzzle properly, participants had to specify the number and gender of the nouns used for naming the puzzle pieces 共e.g., ‘‘Pon los pianos blancos en cuatro, pon la vaca negra en tres,...’’ meaning ‘‘Put the white pianos in four, put the black cow in three,...’’兲. Two native speakers of Spanish independently transcribed the audiotaped responses and coded them for number and gender agreement. A third native speaker resolved any discrepancies between the two transcribers/ coders. Percent agreement between transcribers/coders was greater than 95%. c. Verb-phrase production task. The task was adapted from Curtiss and Yamada’s 共1987兲 CYCLE test to elicit verb morphology 共tense, aspect, person, and number兲 production. Participants heard 20 incomplete sentences, one at a time, illustrated with pictures presented on a computer. They were asked to offer sensible completions. For instance, they might hear ‘‘Ayer fuı´ a la tienda, y yo...’’ 共meaning ‘‘Yesterday I went to the store, and I...’’兲 and see a picture of someone in a store. Next they would see a picture of the person buying milk. They were then given 6 s to complete the sentence. To be counted as an acceptable completion, appropriate morphosyntactic markers had to be used for the verb, as constrained by the lead-in clause 共e.g., first person and singular form in the preterite tense in Spanish in the example just given兲. The 20 items were designed to elicit a variety of tense/aspect, number, and person markings. Agreement between the two transcribers/coders on participants’ audiotaped responses was excellent 共⬎90% agreement兲, and discrepancies were resolved by a third native speaker of Spanish. d. Narrative production. Each frog story produced for the narrative accent rating task 共see Sec. II C b兲 was rated by two native speakers of Spanish on a grammatical wellformedness scale, with 1⫽definitely nonnative and 5 ⫽definitely native. Interrater reliability was excellent 共intraclass R⫽0.90, p⬍0.0001). III. RESULTS A. Phonology 1. Spanish phonemes
a. Prosodic factors. One-way ANOVAs revealed no reliable differences between the three speaker groups in any of Knightly et al.: Production benefits of childhood overhearing
469
TABLE III. Prosodic factors results.
Measure Speech rate 共ms兲 word initial word medial Misplaced stress 共%兲 word initial word medial Phrase boundary before target word 共%兲 word initial word medial
Native speakers
Childhood overhearers
Typical late L2 learners
209 共5兲 195 共6兲
203 共7兲 203 共7兲
196 共7兲 198 共8兲
0.5 共0.4兲 5.6 共2.0兲
1.5 共0.5兲 7.2 共1.9兲
2.2 共1.1兲 10.4 共2.5兲
16.2 共5.8兲 7.8 共2.0兲
25.2 共7.0兲 8.6 共2.7兲
15.7 共5.0兲 7.2 共2.2兲
Note. Standard errors are given in parentheses.
the prosodic factors examined 关speech rate, percentage of misplaced stress, and percentage of prosodic boundary before the target word; F’s(2,42)⬍1.30, n.s.; see Table III兴. These findings suggest that any differences found among the three speaker groups in VOT, percent lenition, and voicing cannot be attributed to these prosodic factors. Rather, the phonetic values seem to reflect the degree of mastery of Spanish phoneme production. b. Voiceless consonants. Childhood overhearers produced word initial /!, #, %/ with shorter VOT 共i.e., more nativelike兲 than did the typical late learners. An ANOVA with place of articulation 共i.e., bilabial/alveolar/velar兲 and consonant position 共i.e., word initial/medial兲 as withinsubject factors and speaker group as a between-subject factor on VOT revealed a main effect of speaker group 关 F(2,42) ⫽6.28, p⬍0.01兴 , place 关 F(2,84)⫽149.91, p⬍0.001兴 , and a reliable position by group interaction 关 F(2,42)⫽8.44, p ⬍0.01兴 . No other reliable main effect or interactions were found. Figure 1 shows the mean VOT for word initial and medial /!, #, %/ for each speaker group. The typical late L2 learners produced reliably longer VOT in word initial position than did both the native speakers and overhearers 共by HSDs, p’s⬍0.01), but the three groups did not differ reliably in VOT for word medial /!, #, %/, thereby yielding the reliable group by position interaction. The main effect of place of articulation was due to longer VOT as the target sounds moved from a bilabial to velar position for all three groups 共by Bonferroni, p’s⬍0.001).
FIG. 1. Mean VOT duration of Spanish word initial and medial /!, #, %/ for each speaker group. 470
J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
FIG. 2. Mean percentage use of lenited consonants in each place of articulation 共i.e., /"/, /$/, /,/兲 for each speaker group.
One-way ANOVAs on speakers’ stop closure duration for initial and medial /!, #, %/ revealed no reliable effects 关 F’s(2,42)⬍2.18, n.s.兴. c. Voiced consonants: degree of lenition. Speakers’ tendency to produce /", $, ,/ as lenited consonants was computed based on the total number of phonetically voiced tokens for a particular phoneme 共e.g., /"/, /$/, or /,/兲 and the number of tokens produced as lenited consonants. An ANOVA was then performed on these lenition percentages with place of articulation and position in word as withinsubject factors, and speaker group as a between-subject factor. There were reliable main effects of speaker group 关 F(2,42)⫽24.03, p⬍0.001兴 , position 关 F(1,42)⫽74.78, p ⬍0.001兴 , and place 关 F(2,84)⫽3.76, p⬍0.05兴 , and a reliable interaction between place and group 关 F(4,84) ⫽5.65, p⬍0.01兴 . Bonferroni posthoc tests showed that all speaker groups produced reliably more lenited consonants in word medial position (M ⫽52.9%, s.e.⫽3.4%) than in word initial position (M ⫽26.4%, s.e.⫽3.5%; p⬍0.001). As can be seen in Fig. 2, the place by group interaction was primarily due to overhearers being less likely to produce /,/ as a lenited consonant than /"/ and /$/ 关 t’s(14)⫽2.65, p’s⬍0.05兴 . This trend was not seen among the native speakers or the typical late L2 learners. Additional follow-up ANOVAs revealed that native speakers outperformed overhearers in producing intervocalic lenited /"/s, who in turn outperformed the typical late L2 learners 关 HSDs, p’s⬍0.05; F(2,42)⫽20.84, p ⬍0.001兴 . For intervocalic /$/, the native speakers and overhearers were comparable, and both outperformed the typical late L2 learners 关 HSDs, p’s⬍0.001; F(2,42)⫽21.91, p ⬍0.001兴 . In contrast, overhearers were no different from the typical late L2 learners in producing intervocalic lenited /,/s, and both did worse than the native speakers 关by HSDs, p’s⬍0.001; F(2,42)⫽18.98, p⬍0.001]. d. Voiced consonants: degree of voicing. Degree of voicing was assessed by the percentage of /", $, ,/ tokens produced by each speaker in each of the three voicing categories 共i.e., voiceless, partial voicing, and full voicing兲, with the total number of /", $, ,/ tokens as the denominator. One-way ANOVAs revealed reliable group differences for partial 关 F(2,42)⫽7.85, p⬍0.01兴 and full voicing 关 F(2,42) ⫽6.40, p⬍0.01兴 but not for voiceless. Tukey posthoc tests Knightly et al.: Production benefits of childhood overhearing
TABLE IV. Accent rating assessment results. Rated speech type Phonemic /!, #, %/ /", $, ,/ Narrative
Native speakers
Childhood overhearers
TABLE V. Participants’ performance on Spanish morphosyntax assessment tasks. Typical late L2 learners Measure
4.4a 共0.08兲 4.4a 共0.10兲 5.0a 共0.0兲
3.6b 共0.08兲 3.4b 共0.10兲 3.0b 共0.23兲
3.0c 共0.09兲 2.8c 共0.10兲 2.4c 共0.17兲
Note. Accent ratings were made using a five-point scale, with higher ratings indicating more nativelike pronunciation. Standard errors are given in parentheses. Within a row, means with different superscripts were reliably different from each other according to Tukey’s HSD test with p⬍0.001 for phonemic ratings, and with p⬍0.05 for narrative ratings.
revealed that native speakers produced /", $, ,/ with full voicing (M ⫽81.7%, s.d.⫽15.2%) more often than typical late L2 learners (M ⫽55.5%, s.d.⫽21.0%; by HSD, p ⬍0.01). Overhearers’ percent use of full voicing (M ⫽70.5%,s.d.⫽23.4%) was in between these two groups, although not reliably different from either. Native speakers used partial voicing less often than typical late L2 learners (M ⫽12.0%, s.d.⫽3.0%; M ⫽29.5%, s.d.⫽3.7%, respectively; by HSD, p⬍0.01). Again, overhearers were in between these two groups but differed from neither reliably (M ⫽13.7%, s.d.⫽3.6%). 2. English phonemes
English production data for word-initial /!/ and /"/ were collected from 14 native speakers, 12 childhood overhearers, and 14 typical late L2 learners.7 One-way ANOVAs revealed no reliable differences among the three groups in speech rate, closure duration, or VOT. All groups produced English /!/ as an aspirated stop with mean VOT values ranging from 50 to 60 ms and English /"/ at 10 ms. The three groups of speakers were also similar in their voicing of English /"/, producing it as a voiceless stop approximately 25% of the time. B. Accent ratings
1. Phonemic accent ratings
Given the excellent interrater agreement 共see Sec. II C a兲, a rating averaged across raters for each speaker was calculated for both voiced and voiceless consonants in word initial and medial position. ANOVAs on these average ratings with group as a between-subject factor and position as a within-subject factor revealed a reliable effect of group for both voiceless 关 F(2,41)⫽71.76, p⬍0.001兴 and voiced consonants 关 F(2,41)⫽72.32, p⬍0.001兴 . No other reliable main effect or interaction was found. Posthoc tests revealed that native speakers received reliably better accent ratings than overhearers, who in turn received reliably better ratings than typical late L2 learners 关by HSDs, p’s⬍0.001; see Table IV兴. Pearson correlations between phonemic accent ratings and phonetic measurements, namely VOT and lenition, were substantial when averaged across tokens 共VOT, r⫽⫺0.53, p⬍0.001; lenition, r⫽0.78, p⬍0.001) as well as at the level of individual tokens 共i.e., second token; VOT, r ⫽⫺0.50, p⬍0.01; lenition, r⫽0.77, p⬍0.001). J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
Native speakers
Grammaticality judgment percentage correct 91.8a 共0.97兲 reaction time 共ms兲 1201a (124) Noun-phrase production gender agreement 94.8a 共2.6兲 number agreement 93.8a 共2.6兲 Verb-phrase production tense/aspect 94.1a 共1.6兲 person 98.2a 共1.0兲 number 98.6a 共1.0兲 Narrative production 4.93a 共6.7兲
Childhood overhearers
Typical late L2 learners
63.6b 共1.6兲 2661b (283)
62.5b 共2.6兲 1936a (307)
66.3b 共6.2兲 82.5a 共5.3兲
72.7b 共2.8兲 92.3a 共1.7兲
50.4b 共5.1兲 68.6b 共4.4兲 70.7b 共4.8兲 2.5b 共0.17兲
50.0b 共5.1兲 72.3b 共5.1兲 80.3b 共4.9兲 2.6b 共0.14兲
Note. Narrative production ratings were on a 5-point scale, with high ratings indicating better formulated sentences. For all other measures, the table indicates the mean percentage correct unless otherwise specified. Standard errors are given in parentheses. Within a row, means with different superscripts were reliably different from each other according to Tukey’s HSD test, p⬍0.01. Numbers with the same superscript were not reliably different from each other.
2. Narrative accent ratings
As shown in Table IV, similar results were found for the frog story accent ratings.8 An ANOVA on the ratings averaged across raters for each speaker revealed reliable group differences 关 F(2,41)⫽73.57, p⬍0.001兴 . Native speakers were rated as having a better whole-sentence accent than the overhearers, who in turn were rated more favorably than the typical late L2 learners 共by Tukey’s HSDs, p’s⬍0.05). C. Morphosyntax
1. Grammaticality judgment
There were reliable group differences for grammaticality judgment both in terms of accuracy and reaction time 关 F(2,41)⫽77.41, p⬍0.001 and F(2,41)⫽8.01, p⬍0.01, respectively; see Table V兴. Native speakers outperformed both childhood overhearers and typical late L2 learners in making correct grammaticality judgments 共Tukey’s HSDs, p’s ⬍0.01). However, the latter two did not differ reliably from each other. The overhearers, but not the typical late L2 learners, took longer than the native speakers in deciding whether a sentence was grammatical 共Tukey’s HSD, p⬍0.01) whereas the latter two did not differ reliably from each other. 2. Noun-phrase production
The native speakers outperformed both the overhearers and the typical late L2 learners in marking gender agreement, but not in marking number agreement 关 F(2,35) ⫽14.38, p⬍0.001 and F(2,35)⫽3.07, p⫽0.06, respectively; see Table V for Tukey’s HSD test results兴. The childhood overhearers and the typical late L2 learners did not differ reliably from each other. 3. Verb-phrase production
For verb morphology production, the native speakers outperformed the childhood overhearers and the typical late Knightly et al.: Production benefits of childhood overhearing
471
L2 learners, who again did not differ reliably from each other 关 F(2,37)⫽9.6, p⬍0.001; see Table V for Tukey’s HSD test results兴. 4. Narrative production
Native speakers’ frog stories were rated as containing fewer morphosyntactic errors than the overhearers’ and the typical late L2 learners’ stories. The latter two did not differ reliably from each other 关 F(2,41)⫽118.27, p⬍0.001; see Table V for Tukey’s HSD test results兴. Taken together, the results of this study reveal considerable benefits of childhood overhearing in phonology and no measurable benefits in morphosyntax. IV. DISCUSSION AND CONCLUSIONS
Childhood overhearers of Spanish were found to have more nativelike Spanish pronunciation than typical late L2 learners according to phonetic analyses 共i.e., VOT, lenition兲, phonemic accent ratings, and narrative accent ratings. Importantly, the overhearers’ advantage in Spanish pronunciation was not found to be attributable to prosodic factors such as speech rate, phrasing of the carrier sentence, and/or the location of stress on the target word. Native speakers, childhood overhearers, and typical late L2 learners were all comparable on these prosodic characteristics. That said, these findings do not necessarily mean that the three groups acquired prosodic features of Spanish to the same degree. Studies on prosodic transfer 共e.g., Ueyama and Jun, 1998; Jun and Oh, 2000兲 show that the degree of proficiency in an L2 is reflected in the realization of L2 intonation patterns and prosodic grouping of words. No reliable group differences were uncovered in the present study perhaps because the sentences evaluated for prosodic factors 共i.e., the target sentences兲 were short and simple. The groups might have shown prosodic differences if we had evaluated production in the storytelling task, which elicited longer and more natural sentences with more complex syntactic structures. Our study speaks to the relation between perception and production. Flege’s speech learning model 共Flege, 1995兲 postulates that a speaker’s pronunciation ability should be only as good as his or her perceptual abilities. When compared to the typical late L2 learners, the childhood overhearers’ better accents may reflect more nativelike perceptual discrimination of Spanish phonemes. Although our study did not test this hypothesis directly, a similar study on Korean L2 acquisition 共Oh et al., 2003兲 suggests that this may indeed be the case. Oh et al. found that adults who regularly heard Korean during childhood performed better in their perception of Korean stops than those who had no exposure to Korean until college. Childhood overhearing, then, may have helped improve speech perception. Such perceptual abilities can be put to use when childhood overhearers try to learn the overheard language later in life 共e.g., in high school and college兲, resulting in a better accent. If this account is on the right track, our findings of childhood overhearing benefits in speech production can be added to the growing body of evidence for a perception-production link. That is, early overhearing trig472
J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
gers better perception, which can be translated into better production even for late language learners 共e.g., Bradlow et al., 1997, 1999兲. However, the logical inference of ‘‘if better perception, then better production’’ needs elaboration. Note that the overhearers in this study, like the typical late L2 learners, rarely produced intervocalic /,/ as a lenited consonant. Nonetheless, they were nativelike for intervocalic /$/, and their intervocalic /"/ fell between these two levels of performance. This pattern of results may be due to the frequency with which the overhearers were exposed to lenited /", $, ,/. For example, lenited /,/ 共i.e., 关$兴兲 is not in the English sound system either phonemically or allophonically, so lack of practice may account for the overhearers’ difficulty in producing lenited /,/. By contrast, the overhearers were nativelike for lenited /$/ 共i.e., 关Z兴兲, which happens to be a highfrequency phoneme in English 共e.g., ‘‘th’’ in this, other兲. For lenited /"/ 共i.e., 关X兴兲, which is not an English phoneme but is close in articulation and similar in voicing to the English /3/, the overhearers produced it better than the typical late learners but not yet nativelike. The puzzle then is why typical late L2 learners had great difficulty producing lenited sounds even when one of them occurs in English 共namely, 关Z兴兲.9 It may have to do with Spanish and English orthography. Since the letters ‘‘b, d, g’’ are produced as stop consonants in English, the typical late L2 learners could be more vulnerable to being misled by the orthographic presentation of Spanish /", $, ,/ than the childhood overhearers. Another possibility is that the typical late L2 learners had not yet acquired the allophones of Spanish /$/ in intervocalic position. That is, even though they knew how to produce the dental fricative, i.e., a lenited /$/, they were not aware that the same sound was also an allophone of /$/ in Spanish. Note that typical late L2 learners also produced Spanish voiced stops with less voicing than did native speakers. Childhood overhearers’ voicing was in between these two groups although not reliably different from either. Combined, these findings hint at the possibility that childhood overhearing of Spanish may lead to better awareness of Spanish allophones. Our findings are also relevant to the nature of input in early childhood bilinguals’ L2 phonology. As in Au et al. 共2002兲, we found that overhearers’ pronunciation advantage did not come at the cost of producing English /!/ and /"/ with a Spanish accent. Nor did the English of our native Spanish speakers 共i.e., Spanish–English bilinguals兲 seem compromised since they also produced English /!/ and /"/ with values comparable to that of typical late L2 learners. These findings are consistent with prior findings where Spanish–English bilinguals who learned English by age 5– 6 produced English /#/ 共Flege, 1991兲, /!/ 共Williams, 1980兲, and /"/ 共Flege and Eefting, 1988兲 like English monolinguals. However, our results contrast with findings where Spanish– English bilinguals produced English /!, #, %/ with ‘‘compromised’’ VOT values 共Flege and Eefting, 1987兲 and English /"/ with Spanish-like voicing lead 共Flege and Eefting, 1988兲. One way to make sense of these conflicting results is to focus on the nature of English input. Flege and Eefting 共1987兲 speculated that the early childhood bilinguals in their Knightly et al.: Production benefits of childhood overhearing
study had ‘‘compromised’’ VOT values because much of their English input was Spanish-accented; they were born in the U.S. but were attending a university in a Spanishspeaking country at the time of testing. In contrast, the bilinguals in Williams 共1980兲 and Flege 共1991兲 were more like the Spanish–English bilinguals in our study in that they were living in the U.S. at the time of testing and used both languages in their daily lives. Thus, the nature 共hearing good models of American English兲 as well as the timing of L2 input may play an important role in childhood bilinguals’ eventual L2 pronunciation. An important conclusion of the current study is that the childhood overhearing advantage seems to be domain specific: it is very robust in phonology but so far not detectable in morphosyntax—a domain, like phonology, that is easy for children to acquire but difficult for adults to master. Even with our rather comprehensive assessment of morphosyntax with additional tasks not included in Au et al.’s 共2002兲 study 共i.e., elicited verb-phrase production, story production兲, no childhood overhearing advantage in morphosyntax was detected. Perhaps it takes more than merely overhearing a language during childhood to gain an edge in morphosyntax acquisition. Future research can help determine whether other kinds of childhood language experience, such as speaking a language for a few years during childhood, results in measurable benefits in mastery of morphosyntax as well as phonology for adults 共re-兲learning a childhood language. Our results can have important applied implications for immigrant children in a predominantly monolingual nation such as the U.S. Most immigrant children in such linguistic environments tend to lose, or severely limit their use of, their heritage language 共e.g., Fillmore, 1991; Au, in press兲. Our study can help immigrant parents and early childhood educators make more informed decisions about what kind of childhood language environment and experience they want to foster for their children. Specifically, this study suggests that overhearing a language during childhood can probably translate into a measurable accent advantage when children try to acquire the overheard language later in life. ACKNOWLEDGMENTS
We thank Celia Adame, Marilyn Alvarado, Brenda Alvarez, Argelia Andrade, Cecilia Arriaza, J. Patrick Barjam, Jessica Brauner, Mauricio Carvallo, Ivonne Cisneros, Susan ˜ a, Curtiss, Kristina Cutura, Olga Escamilla, Laura Espan Catherine Fountain, Ingrid Gooding, John Grinstead, Vilma Hernandez, Gerardo Ibarra, Patricia Keating, Sahyang Kim, Hyuck-Joon Lee, Reuben Lim, Ann Marroquin, Olivia Martı´nez, Karla Mayen, Nancy Miranda, Claudia Parodi, Jaime Paz, Katie Polsky, Irasema Ramos, Salome Rinco´n, Krista Rodriguez-Bruno, Laura Romo, Norma Salazar, Susan Schaffer, Sylvia Sherno, Minjung Son, Henry Tehrani, Motoko Ueyama, and John Whalen for their help with various aspects of this research. We thank the research participants for their time. We also thank Anders Lo¨fqvist and two anonymous reviewers for their insightful and valuable comments on an earlier version of this paper. This work was supported by NIMH Grant No. MH56118 共Au and Jun兲, an NIMH supplement 共Au and Knightly兲, an NSF predoctoral fellowJ. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
ship 共Oh兲, NIMH traineeships in Developmental Cognitive Science 5T32MH10926 共Knightly, Oh兲, and seed grants from University of California Linguistic Minority Research Institute, UCLA Institute of American Cultures, and UCLA Academic Senate 共Au兲. Among the 45 speakers participating in the current study, 18 speakers 共5 of the 15 native speakers, 6 of the 15 childhood overhearers, and 7 of the 15 typical late L2 learners兲 contributed data to Au et al.’s 共2002兲 interim report, which also included 15 other speakers from a pilot study. 共i.e., 5 natives, 5 childhood overhearers, and 5 typical late L2 learners兲. 2 To assess inter-measurer/coder agreement, the primary measurer/coder analyzed the Spanish data on two speakers measured by each of the two secondary measurers/coders, and the English data on two speakers measured by the only secondary measurer/coder. Thus, the reported ranges of agreement are based on the four values for Spanish and two values for English. 3 Except where noted, all native Spanish-speaking raters were of Mexican or Central American descent. In addition, speech samples assessed by two native Spanish speakers 共Mexican兲 verified self-reports of raters’ native speaker status. 4 Of the 48 raters, 36 were of Mexican or Central American descent, 5 were from Spain, and 7 were of South American descent; 3 from Argentina and 1 each from Columbia, Peru, Ecuador and Bolivia. 5 Even though raters were instructed to focus on the target segment, their ratings could be influenced by the carrier sentence. However, since the carrier sentence was short and repeated for every target sentence, its effect may be minimal. Compared to narrative accent ratings based on more varied and longer sentences, the phonemic accent ratings are likely to reflect the quality of the target phonemes. 6 Since the narrative accent ratings involved a much smaller data set 共45 story sessions兲 compared to the phonemic accent ratings, only two raters of Mexican American descent, who readily achieved excellent interrater reliability, were recruited for this task. 7 One native speaker, three overhearers, and one typical late L2 learner were not available for testing when English data were collected about two months after the Spanish data collection. 8 The relatively low phonemic accent ratings for native speakers may reflect the difficulty in rating just the target sounds embedded in the carrier sentence. The raters may therefore have been more conservative, using a narrower range of the rating scale and staying closer to the mid-point. On the other hand, raters may have been more confident about their narrative accent ratings based on the longer speech samples 共i.e., stories兲 and hence showed more range in their ratings. 9 Examination of individual lenition percentages revealed that two typical late L2 learners failed to produce any lenited consonants. 1
Au, T. K. 共in press兲. ‘‘Salving heritage languages,’’ in Heritage Language Acquisition: A New Field Emerging, edited by D. Brinton and O. Kagan 共Erlbaum, Mahwah, NJ兲. Au, T. K., Knightly, L. M., Jun, S., and Oh, J. S. 共2002兲. ‘‘Overhearing a language during childhood,’’ Psychol. Sci. 13, 238 –243. Berman, R., Slobin, D. I., Aksu-Koc, A. A., Bamberg, M., Dasinger, L., Marchman, V., Neeman, Y., Rodkin, P. C., and Sebastian, E. 共1994兲. Relating Events in Narrative: A Cross Linguistic Developmental Study 共Erlbaum, Hillsdale, NJ兲. Birdsong, D., and Molis, M. 共2001兲. ‘‘On the evidence for maturational constraints in second-language acquisition,’’ J. Memory Lang. 44, 235– 249. Bongaerts, T., Van Summeren, C., Planken, B., and Schils, E. 共1997兲. ‘‘Age and ultimate attainment in the pronunciation of a foreign language,’’ Studies in Second Lang. Acquis. 19, 447– 465. Bradlow, A. R., Akahane-Yamada, R., Pisoni, D. B., and Tohkura, Y. 共1999兲. ‘‘Training Japanese listeners to identify English /./ and /(/: Long-term retention of learning in perception and production,’’ Percept. Psychophys. 61, 977–985. Bradlow, A. R., Pisoni, D. B., Akahane-Yamada, R., and Tohkura, Y. 共1997兲. ‘‘Training Japanese listeners to identify English /./ and /(/: IV. Some effects of perceptual learning on speech production,’’ J. Acoust. Soc. Am. 101, 2299–2310. Cohen, J. D., MacWhinney, B., Flatt, M., and Provost, J. 共1993兲. ‘‘PsyScope: A new graphic interactive environment for designing psychology Knightly et al.: Production benefits of childhood overhearing
473
experiments,’’ Behav. Res. Methods Instrum. Comput. 25, 257–271. Curtiss, S., and Yamada, J. 共1987兲. ‘‘The Curtiss-Yamada Comprehensive Language Evaluation 共CYCLE兲,’’ University of California, Los Angeles. de Jong, K. 共1995兲. ‘‘The supraglottal articulation of prominence in English: Linguistic stress as localized hyperarticulation,’’ J. Acoust. Soc. Am. 97, 491–504. Fillmore, L. W. 共1991兲. ‘‘When learning a second language means losing the first,’’ Early Childhood Res. Qu. 6, 323–346. Flege, J. E. 共1991兲. ‘‘Age of learning affects the authenticity of voice-onset time 共VOT兲 in stop consonants produced in a second language,’’ J. Acoust. Soc. Am. 89, 395– 411. Flege, J. E. 共1995兲. ‘‘Second language speech learning: Theory, findings, and problems,’’ in Speech Perception and Linguistic Experience: Theoretical and Methodological Issues in Cross-language Speech Perception, edited by W. Strange 共York, Baltimore, MD兲, pp. 233–272. Flege, J. E., and Eefting, W. 共1987兲. ‘‘Production and perception of English stops by native Spanish speakers,’’ J. Phonetics 15, 67– 83. Flege, J. E., and Eefting, W. 共1988兲. ‘‘Imitation of a VOT continuum by native speakers of English and Spanish: Evidence for phonetic category formation,’’ J. Acoust. Soc. Am. 83, 729–740. Flege, J. E., and Fletcher, K. L. 共1992兲. ‘‘Talker and listener effects on degree of perceived foreign accent,’’ J. Acoust. Soc. Am. 91, 370–389. Flege, J. E., Munro, M. J., and MacKay, I. R. A. 共1995兲. ‘‘Factors affecting strength of perceived foreign accent in a second language,’’ J. Acoust. Soc. Am. 97, 3125–3134. Flege, J. E., Yeni-Komshian, G. H., and Liu, S. 共1999兲. ‘‘Age constraints on second-language acquisition,’’ J. Memory Lang. 41, 78 –104. Fougeron, C., and Keating, P. A. 共1997兲. ‘‘Articulatory strengthening at edges of prosodic domains,’’ J. Acoust. Soc. Am. 101, 3728 –3740. Johnson, E. K., and Jusczyk, P. W. 共2001兲. ‘‘Word segmentation by 8-montholds: When speech cues count more than statistics,’’ J. Memory Lang. 44, 548 –567. Johnson, J. S., and Newport, E. L. 共1989兲. ‘‘Critical period effects in second language learning: The influence of maturational state on the acquisition of English as a second language,’’ Cognit Psychol. 21, 60–99. Johnson, J. S., Shenkman, K. D., Newport, E. L., and Medin, D. L. 共1996兲. ‘‘Indeterminacy in the grammar of adult language learners,’’ J. Memory Lang. 35, 335–352. Jun, S.-A., and Oh, M. 共2000兲. ‘‘Acquisition of 2nd Language Intonation,’’ in Proceedings of International Conference on Spoken Language Processing 共Beijing, China兲, Vol. 4, pp. 76 –79. Kuhl, P. K., Williams, K. A., Lacerda, F., Stevens, K. N., and Lindblom, B. 共1992兲. ‘‘Linguistic experience alters phonetic perception in infants by 6 months of age,’’ Science 共Washington, DC, U.S.兲 255, 606 – 608. Lisker, L., and Abramson, A. 共1964兲. ‘‘A cross-language study of voicing in initial stops: Acoustical measurements,’’ Word 20, 384 – 422. Oh, J. S., Jun, S., Knightly, L. M., and Au, T. K. 共2003兲. ‘‘Holding on to childhood language memory,’’ Cognition 86, B53–B64.
474
J. Acoust. Soc. Am., Vol. 114, No. 1, July 2003
Oyama, S. 共1976兲. ‘‘A sensitive period for the acquisition of a nonnative phonological system,’’ J. Psycholinguist. Res. 5, 261–283. Patkowski, M. 共1980兲. ‘‘The sensitive period for the acquisition of syntax in second language,’’ Lang. Learning 30, 449– 472. Phinney, J. S. 共1992兲. ‘‘The multigroup ethnic identity measure: a new scale for use with diverse groups,’’ J. Adolescent Res. 7, 156 –176. Plann, S. J. 共1979兲. ‘‘Morphological problems in the acquisition of Spanish in an immersion classroom,’’ in The Acquisition and Use of Spanish and English as First and Second Languages, edited by R. W. Anderson 共TESOL, Washington, DC兲, pp. 119–132. Saffran, J. R., Aslin, R. N., and Newport, E. L. 共1996兲. ‘‘Statistical learning by 8-month-old infants,’’ Science 共Washington, DC, U.S.兲 274, 1926 – 1928. Saffran, J. R., Newport, E. L., Aslin, R. N., Tunick, R. A., and Barrueco, S. 共1997兲. ‘‘Incidental language learning: Listening 共and Learning兲 out of the corner of your ear,’’ Psychol. Sci. 8, 101–105. Snow, C. E., and Hoefnagel-Hohle, M. 共1978兲. ‘‘The critical period for language acquisition: evidence from second language learning,’’ Child Dev. 49, 1114 –1128. Stockwell, R. P., and Bowen, J. D. 共1965兲. The Sounds of English and Spanish 共Univ. of Chicago, Chicago兲. Tahta, S., Wood, M., and Loewenthal, K. 共1981兲. ‘‘Foreign accents: Factors relating to transfer of accent from the first language to a second language,’’ Lang. Speech 24, 265–272. Tees, R. C., and Werker, J. F. 共1984兲. ‘‘Perceptual flexibility: Maintenance or recovery of the ability to discriminate non-native speech sounds,’’ Can. J. Psychol. 38, 579–590. Ueyama, M., and Jun, S.-A. 共1998兲. ‘‘Focus realization in Japanese English and Korean English intonation,’’ in Japanese and Korean Linguistics 共CSLI, Cambridge U.P., Cambridge兲, Vol. 7, pp. 629– 645. Werker, J. F. 共1995兲. ‘‘Age-related changes in cross-language speech perception: Standing at the crossroads,’’ in Speech Perception and Linguistic Experience: Theoretical and Methodological Issues in Cross-language Speech Perception, edited by W. Strange 共York, Baltimore, MD兲, pp. 155– 169. Werker, J. F., and Tees, R. C. 共1984兲. ‘‘Cross-language speech perception: Evidence for perceptual reorganization during the first year of life,’’ Infant Behav. Dev. 7, 49– 63. Williams, L. 共1980兲. ‘‘Phonetic variation as a function of second-language learning,’’ in Child Phonology Volume 2: Perception, edited by G. H. Yeni-Komshian, J. F. Kavanagh, and C. A. Ferguson 共Academic, New York兲, pp. 185–215. Wode, H. 共1981兲. Learning a Second Language: An Integrated View of Language Acquisition 共Gunter Narr Verlag, Tubingen兲. Yamada, R. A. 共1995兲. ‘‘Age and acquisition of second language speech sounds: perception of American English /./ and /(/ by native speakers of Japanese,’’ in Speech Perception and Linguistic Experience: Theoretical and Methodological Issues, edited by W. Strange 共York, Baltimore兲, pp. 305–320.
Knightly et al.: Production benefits of childhood overhearing